Chinese Journal of Lasers, Volume. 48, Issue 9, 0910001(2021)
Hyperspectral Image Classification Method Based on Image Reconstruction Feature Fusion
Figures & Tables(12)
Fig. 2. Spatial domain blocks of pixel xij. (a) Normal position; (b) edge position; (c) corner position
Fig. 4. Indian Pines hyperspectral image. (a) False colour image; (b) actual feature map
Fig. 5. Pavia University hyperspectral image. (a) False colour image; (b) actual feature map
Fig. 7. Classification results of each methods on Indian Pines dataset. (a) False colour image; (b) ground truth; (c) KNN method; (d) SAM method; (e) SVM method; (f) EPF method; (g) LBP-SVM method; (h) SVMCK method; (i) LBP-SAM method; (j) CDSRC method; (k) CCJSR method; (l) RSFM method
Fig. 8. Classification results of each methods on Pavia University dataset. (a) False colour image; (b) ground truth; (c) KNN method; (d) SAM method; (e) SVM method; (f) EPF method; (g) LBP-SVM method; (h) SVMCK method; (i) LBP-SAM method; (j) CDSRC method; (k) CCJSR method; (l) RSFM method
Table 1. Classification results of different methods on Indian Pines dataset
View tableTable 1. Classification results of different methods on Indian Pines dataset
Method (OA±Std) /% A=1% A=2% A=4% A=6% A=8% A=10% KNN 55.08±1.44 59.30±1.43 63.54±1.03 66.38±0.47 67.62±0.67 68.61±0.60 SAM 54.53±2.07 60.01±1.10 63.76±0.84 66.04±0.66 68.11±0.87 69.03±0.48 SVM 56.98±2.16 65.23±1.49 72.29±1.29 76.25±1.41 79.22±0.78 80.71±0.72 EPF 69.40±3.50 76.51±4.23 87.37±2.50 91.31±1.92 93.24±1.49 94.93±0.89 LBP-SVM 78.37±2.08 84.76±2.01 91.48±0.72 93.13±0.45 94.76±0.32 96.61±0.34 SVMCK 80.46±1.67 84.94±1.78 91.56±0.45 94.50±0.72 96.08±0.42 96.94±0.52 LBP-SAM 74.08±2.77 82.33±2.01 88.19±1.61 92.05±1.28 94.82±1.15 95.12±0.68 CDSRC 73.84±2.30 77.63±0.90 79.62±0.75 80.80±0.53 81.22±0.89 82.18±0.41 CCJSR 70.55±2.32 79.67±1.44 87.16±1.01 91.78±0.52 94.38±0.71 95.65±0.42 RSFM 86.87±1.78 93.26±1.11 96.85±0.77 98.12±0.39 98.84±0.36 99.06±0.31
Table 2. Classification results of various ground objects in Indian Pines dataset under different methods
View tableTable 2. Classification results of various ground objects in Indian Pines dataset under different methods
Class KNN SAM SVM EPF LBP-SVM SVMCK LBP-SAM CDSRC CCJSR RSFM 1 0.346 0.735 0.683 0.585 0.975 1.000 0.090 0.892 1.000 1.000 2 0.538 0.547 0.699 0.922 0.966 0.945 0.929 0.752 0.956 0.995 3 0.510 0.632 0.762 0.933 0.929 0.941 0.920 0.707 0.927 0.979 4 0.375 0.442 0.686 0.873 0.953 0.830 0.778 0.678 0.898 0.972 5 0.801 0.844 0.799 0.963 0.959 0.984 0.995 0.911 0.972 0.984 6 0.825 0.826 0.916 1.000 0.973 0.991 0.996 0.941 0.969 0.999 7 0.700 0.750 0.700 0.957 0.821 1.000 0.850 0.917 0.793 1.000 8 0.907 0.945 0.967 1.000 1.000 1.000 0.973 0.982 0.995 1.000 9 0.314 0.333 0.483 1.000 0.938 0.667 1.000 0.833 0.938 1.000 10 0.602 0.583 0.719 0.857 0.974 0.952 0.856 0.753 0.964 0.979 11 0.691 0.721 0.771 0.966 0.968 0.973 0.947 0.761 0.959 0.994 12 0.491 0.479 0.808 0.961 0.924 0.998 0.972 0.793 0.917 0.989 13 0.861 0.775 0.879 0.995 0.977 1.000 0.957 0.973 0.974 0.984 14 0.906 0.891 0.942 0.997 1.000 0.969 0.999 0.937 0.991 1.000 15 0.502 0.498 0.623 0.891 0.949 0.987 0.913 0.725 0.959 1.000 16 0.988 0.987 0.964 1.000 0.960 1.000 0.924 0.914 0.963 1.000 OA 0.682 0.697 0.795 0.948 0.966 0.964 0.951 0.813 0.959 0.992 AA 0.647 0.685 0.775 0.931 0.954 0.957 0.946 0.842 0.948 0.971 κ 0.637 0.654 0.765 0.941 0.961 0.959 0.943 0.786 0.953 0.991
Table 3. Classification results of different classification methods on Pavia University dataset
View tableTable 3. Classification results of different classification methods on Pavia University dataset
Method (OA±Std) /% A=1% A=2% A=4% A=6% A=8% A=10% KNN 77.79±0.71 79.88±0.48 81.79±0.47 82.52±0.35 83.18±0.19 83.62±0.18 SAM 77.80±0.64 79.82±0.52 81.75±0.45 82.57±0.22 83.11±0.41 83.60±0.20 SVM 84.97±0.63 87.97±0.67 89.47±0.30 90.39±0.34 90.90±0.11 91.12±0.20 EPF 93.11±1.63 95.16±1.26 96.24±0.81 97.18±0.61 97.86±0.40 98.01±0.24 LBP-SVM 90.78±0.76 94.61±0.42 96.11±0.29 97.03±0.31 97.65±0.27 97.97±0.13 SVMCK 92.54±3.34 96.21±0.53 97.86±0.32 98.34±0.26 98.75±0.07 98.91±0.15 LBP-SAM 91.74±1.18 92.67±1.24 94.04±0.81 95.69±0.71 96.11±0.53 96.32±0.42 CDSRC 81.04±0.89 82.63±0.53 83.79±0.28 84.21±0.42 84.58±0.25 85.03±0.23 CCJSR 78.17±0.67 82.95±0.51 87.80±0.40 90.26±0.36 92.33±0.21 93.40±0.27 RSFM 96.51±0.59 98.38±0.49 99.13±0.30 99.44±0.24 99.58±0.18 99.73±0.16
Table 4. Classification results of various ground objects in Pavia University dataset under different methods
View tableTable 4. Classification results of various ground objects in Pavia University dataset under different methods
Class KNN SAM SVM EPF LBP-SVM SVMCK LBP-SAM CDSRC CCJSR RSFM 1 0.918 0.917 0.913 0.994 0.966 0.999 0.997 0.891 0.952 1.000 2 0.872 0.874 0.939 1.000 0.994 0.987 0.993 0.884 0.975 0.998 3 0.627 0.647 0.861 0.774 0.970 0.989 0.941 0.724 0.835 1.000 4 0.961 0.946 0.959 0.984 0.941 1.000 0.970 0.942 0.990 0.999 5 0.989 0.998 0.997 1.000 0.920 1.000 0.999 0.997 0.947 1.000 6 0.705 0.710 0.866 0.951 0.998 0.977 0.906 0.709 0.928 0.999 7 0.690 0.671 0.864 1.000 0.975 0.998 0.960 0.764 0.879 1.000 8 0.655 0.654 0.787 0.999 0.988 0.988 0.949 0.717 0.806 1.000 9 0.998 0.996 0.993 0.999 0.839 0.997 1.000 1.000 0.856 1.000 OA 0.835 0.835 0.912 0.981 0.979 0.910 0.965 0.851 0.937 0.998 AA 0.824 0.824 0.909 0.967 0.955 0.993 0.958 0.847 0.908 0.999 κ 0.778 0.778 0.882 0.975 0.971 0.986 0.956 0.801 0.916 0.998
Tools
Get Citation
Copy Citation Text
Jiamin Liu, Chao Zheng, Limei Zhang, Zehua Zou. Hyperspectral Image Classification Method Based on Image Reconstruction Feature Fusion[J]. Chinese Journal of Lasers, 2021, 48(9): 0910001
Paper Information
Category: remote sensing and sensor
Received: Sep. 4, 2020
Accepted: Nov. 18, 2020
Published Online: May. 17, 2021
The Author Email: Liu Jiamin (liujm@cqu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20